Granite Embedding Multilingual R2: Open Apache 2.0 Multilingual Embeddings with 32K Context — Best Sub-100M Retrieval Quality

IBM released Granite Embedding Multilingual R2, two new multilingual embedding models balancing model size with retrieval quality.

• •97M compact model achieves 60.3 on MTEB Multilingual Retrieval, the highest for open sub-100M models, with 9.4 point lead over peers.
• •311M full-size model scores 65.2 on MTEB Multilingual Retrieval, #2 among open models under 500M parameters with Matryoshka support.
• •Both support 200+ languages with enhanced training for 52 languages, 32K-token context (64x R1), and code retrieval in 9 programming languages.
• •Available under Apache 2.0 license with ONNX and OpenVINO weights, compatible with LangChain, LlamaIndex, Haystack, and Milvus.

This summary was automatically generated by AI based on the original article and may not be fully accurate.